🤖 AI Code Review - GitHub Action with Deep Static Analysis

Automated AI-powered code review for GitHub Pull Requests
GPT-5 • Claude Opus 4 • Gemini 2.5 Pro • O3 • Custom AI Models

🚀 Quick Start • ✨ Features • 📖 Documentation • 💡 Examples • 🤝 Contributing

---

🎯 What is AI Code Review?

AI Code Review is a next-generation GitHub Action that automates pull request reviews using advanced artificial intelligence and deep static code analysis. Get Principal/Staff engineer-level feedback on every PR with AST parsing, linter integration, dependency tracking, and beautiful visual statistics.

🏆 Why Choose AI Code Review?

✅ Save 2-4 hours per day on code reviews
✅ Catch bugs before production with deep analysis
✅ Consistent quality standards across your team
✅ Support for 10+ programming languages
✅ Works with GPT-5, Claude 4, Gemini, O3, and custom AI models
✅ AST parsing + Linter integration + Dependency analysis

---

✨ Key Features

🧠 Senior-Level AI Reviews

• AI analyzes code like a Principal engineer with 15+ years of experience
• Deep understanding of architecture, performance, and security
• Provides detailed "why" explanations, not just "what"
• Recognizes good code and best practices

🔍 Deep Static Code Analysis

• AST Parsing: Extract functions, classes, and dependencies from any language
• Linter Integration: Auto-runs ESLint, Pylint, Rust Clippy, C# Analyzers
• Dependency Tracking: Maps function calls and dependencies across codebase
• Complexity Metrics: Cyclomatic complexity and maintainability index
• Call Graph Analysis: Understand function relationships and breaking changes

🎨 Beautiful Visual Reports

┌─────────────────────────────────────────────────────────────┐
│ Files Reviewed    │ 15                                      │
│ Quality Score     │ ████████████████░░░░ 85/100             │
│ Issues Found      │ Critical: 0 | Warnings: 2 | Info: 1     │
└─────────────────────────────────────────────────────────────┘

• ASCII charts and sparklines
• Quality scores and visual statistics
• Trend analysis over time

🤖 Multi-Model Support

• OpenAI: GPT-5, GPT-5-High, O3, O3-Mini
• Anthropic: Claude Opus 4, Claude Sonnet 4
• Google: Gemini 2.5 Pro, Gemini 2.5 Flash
• Custom Endpoints: Azure OpenAI, AWS Bedrock, local models, or any OpenAI-compatible API

🛠️ AI Tool System (14 Tools)

AI actively investigates your code using specialized tools:

• read_file, get_file_diff, analyze_file_ast - Deep file analysis
• analyze_function_impact - 🎯 Breaking change detector! Shows full context around ALL function call sites
• find_function_callers, find_function_dependencies - Code navigation
• analyze_function_complexity - Complexity metrics
• run_linter - Execute language-specific linters
• get_commits_list, get_commit_diff, get_file_history - Git context
• ...and more!

🌍 Multi-Language Support

• Review Language: Comments in any language (English, Russian, Spanish, French, German, Chinese, etc.)
• Code Languages: TypeScript, JavaScript, Python, Rust, C#, Go, Java, PHP, Ruby, and more

📦 Smart PR Handling

• Intelligent Chunking: Splits large PRs optimally by module boundaries
• Handles massive PRs without hitting token limits
• Silent Mode: Reduce notification spam for your team
• Auto Labeling: Manages PR labels based on review results

---

🚀 Quick Start

Step 1: Create Workflow File

Create .github/workflows/ai-code-review.yml:

name: AI Code Review

on:
  pull_request:
    types: [opened, synchronize, reopened]

jobs:
  ai-review:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write
    
    steps:
      - name: Checkout code
        uses: actions/checkout@v4
        with:
          fetch-depth: 0  # Important for commit history analysis

      - name: Run AI Code Review
        uses: zxcloli666/AI-Code-Review@v1
        with:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Step 2: Add OpenAI API Key

1. Get your API key from OpenAI Platform
2. Go to your repository Settings → Secrets and variables → Actions
3. Click New repository secret
4. Name: OPENAI_API_KEY, Value: your-api-key-here

Step 3: Create a Pull Request

That's it! 🎉 AI Code Review will automatically analyze every new PR.

---

📖 Configuration Options

Required Inputs

Input	Description	Default
GITHUB_TOKEN	GitHub token (auto-provided by Actions)	-
OPENAI_API_KEY	OpenAI API key or compatible provider	-

Optional Inputs

Input	Description	Default
OPENAI_API_MODEL	AI model (gpt-5, claude-opus-4, o3, etc.)	gpt-5
OPENAI_API_BASE_URL	Custom endpoint (Azure, Bedrock, local)	https://api.openai.com/v1
REVIEW_LANGUAGE	Review language (en, ru, es, fr, de, zh, etc.)	en
SILENT_MODE	Minimize email notifications (true/false)	false
MAX_CHUNK_SIZE	Max tokens per chunk (adjust for your model)	6000
ENABLE_LINTERS	Run language-specific linters	true
ENABLE_AST	AST analysis for deep code understanding	true
ENABLE_DEPENDENCY_ANALYSIS	Dependency tracking and call graphs	true
SEVERITY_THRESHOLD	Min severity to report (info, warning, error)	warning

---

💡 Example Reviews

🔍 Click to see example review output

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃                                                                   ┃
┃      🤖 𝗔𝗜 𝗖𝗢𝗗𝗘 𝗥𝗘𝗩𝗜𝗘𝗪 - 𝗔𝗡𝗔𝗟𝗬𝗦𝗜𝗦 𝗖𝗢𝗠𝗣𝗟𝗘𝗧𝗘 🤖              ┃
┃                                                                   ┃
┃         ⚡ Powered by Advanced AI & Deep Code Analysis ⚡           ┃
┃                                                                   ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

## 📋 Executive Summary

This PR implements a new caching layer to reduce database load and improve 
response times. The implementation is solid with good error handling and test 
coverage. Found 2 performance optimizations and 1 documentation suggestion.

**Verdict**: ✅ APPROVED with minor suggestions

## 📊 Review Overview

┌─────────────────────────────────────────────────────────────┐
│ Files Reviewed        │ 15                                  │
│ Total Lines Changed   │ 450 (+320/-130)                     │
│ Commits Analyzed      │ 8                                   │
│ Tools Used            │ 12                                  │
│ Review Time           │ 45s                                 │
│ Quality Score         │ ████████████████░░░░ 85/100         │
└─────────────────────────────────────────────────────────────┘

## 🎯 Issues Summary

Critical  ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░  0
Warnings  ████████████░░░░░░░░░░░░░░░░░░░░░░░░░░░░  2
Info      ████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░  1

## 🏆 Highlights

✅ Well-structured code with clear separation of concerns
✅ Comprehensive error handling in API endpoints
✅ Good use of TypeScript types for type safety

## ⚠️ Issues Found

### src/api/handler.ts:45
**Severity:** Warning | **Category:** Performance

The database query inside the loop creates N+1 queries. Consider using 
a single query with JOIN or batch loading.

**Current:**
\`\`\`typescript
for (const user of users) {
  const posts = await db.query('SELECT * FROM posts WHERE user_id = ?', [user.id]);
}
\`\`\`

**Suggested:**
\`\`\`typescript
const posts = await db.query(`
  SELECT p.*, u.id as user_id
  FROM posts p
  JOIN users u ON p.user_id = u.id
  WHERE u.id IN (?)
`, [users.map(u => u.id)]);
\`\`\`

---

🎯 Advanced Use Cases

Manual Trigger with /review Command

Add the ability to trigger reviews on-demand:

on:
  pull_request:
    types: [opened, synchronize, reopened]
  issue_comment:
    types: [created]

jobs:
  ai-review:
    if: |
      (github.event_name == 'pull_request') ||
      (github.event_name == 'issue_comment' &&
       github.event.issue.pull_request &&
       contains(github.event.comment.body, '/review'))
    # ... rest of the workflow

Now comment /review on any PR to trigger a review!

Use Claude Opus 4

- uses: zxcloli666/AI-Code-Review@v1
  with:
    GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
    OPENAI_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
    OPENAI_API_BASE_URL: 'https://api.anthropic.com/v1'
    OPENAI_API_MODEL: 'claude-opus-4'

Use Google Gemini 2.5 Pro

- uses: zxcloli666/AI-Code-Review@v1
  with:
    GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
    OPENAI_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
    OPENAI_API_BASE_URL: 'https://generativelanguage.googleapis.com/v1'
    OPENAI_API_MODEL: 'gemini-2.5-pro'
    MAX_CHUNK_SIZE: '12000'  # Gemini supports larger context

Use OpenAI O3 Reasoning Model

- uses: zxcloli666/AI-Code-Review@v1
  with:
    GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
    OPENAI_API_MODEL: 'o3-mini'
    MAX_CHUNK_SIZE: '10000'

Reviews in Russian

- uses: zxcloli666/AI-Code-Review@v1
  with:
    GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
    REVIEW_LANGUAGE: 'ru'  # Reviews in Russian

Silent Mode (Reduce Notifications)

- uses: zxcloli666/AI-Code-Review@v1
  with:
    GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
    SILENT_MODE: 'true'  # Minimize email spam

---

🛠️ How It Works

1️⃣ Data Collection

• Retrieves PR details, branch info, and linked issues
• Analyzes file types and programming languages
• Builds commit history timeline

2️⃣ Smart Chunking

• Splits large PRs intelligently by module boundaries
• Groups related files together
• Optimizes token usage for your chosen AI model

3️⃣ Deep Analysis

AI actively investigates code using 14 specialized tools:

File Analysis:

• read_file - Read complete file content
• get_file_diff - View specific changes with context
• analyze_file_ast - Deep AST (Abstract Syntax Tree) parsing

Code Understanding:

• analyze_function_impact - 🎯 Breaking Change Detector! Shows full context around ALL function call sites
• find_function_callers - List all places where a function is called
• find_function_dependencies - Find what a function depends on
• analyze_function_complexity - Cyclomatic complexity metrics

Quality Tools:

• run_linter - Execute language-specific linters (ESLint, Pylint, Clippy, etc.)

Git & History:

• get_commits_list, get_commit_diff, get_file_history - Analyze code evolution

4️⃣ AI Review Process

• Uses mandatory workflow requiring 4-6 tools minimum per review
• Analyzes from multiple dimensions: Security, Performance, Architecture, Maintainability
• Provides senior-level feedback with detailed explanations
• Categorizes issues by severity (critical, warning, info)

5️⃣ Executive Summary Generation

• Sends full review to AI for concise summary
• Creates brief but comprehensive executive summary (under 300 words)
• Includes: main changes, critical findings, key recommendations, verdict

6️⃣ Publishing Results

• Posts comprehensive review comment
• Adds inline comments on specific lines
• Manages labels automatically
• Supports silent mode to reduce notification spam

---

🔥 Breaking Change Detection Example

When refactoring a function used across multiple files, AI uses analyze_function_impact to show full context around ALL call sites.

Example: Changing calculatePrice(quantity, price) to calculatePrice(options)

// Before:
function calculatePrice(quantity: number, price: number): number {
  return quantity * price;
}

// After:
function calculatePrice(options: { quantity: number; price: number; discount?: number }): number {
  const { quantity, price, discount = 0 } = options;
  return quantity * price * (1 - discount);
}

What AI sees:

## Function Impact Analysis: calculatePrice

**Definition**: `src/utils/pricing.ts:42`
**Parameters**: quantity, price
**Exported**: Yes

### Call Sites

Found **8** call site(s) across **3** file(s)

---

#### `src/components/Cart.tsx` (3 calls)

**Call at line 67**:
→ 67 | const total = calculatePrice(item.quantity, item.price);
  68 | setCartTotal(total);

**Call at line 89**:
  88 | items.forEach(item => {
→ 89 |   subtotal += calculatePrice(item.qty, item.unitPrice);
  90 | });

---

Total Impact: 8 call site(s) would be affected

Recommendations:
⚠️ Medium impact: 8 call sites - thorough testing recommended
- All callers pass 2 parameters - breaking change confirmed
- Consider backward compatibility or deprecation period

The AI can now intelligently warn that all 8 call sites need updating! 🎯

---

🌐 Supported Languages

Review Languages (Comments)

English, Russian, Spanish, French, German, Chinese, Japanese, Korean, Portuguese, Italian, Dutch, Polish, Turkish, Arabic, Hindi, and more...

Code Analysis Languages

• JavaScript/TypeScript - ESLint, AST parsing
• Python - Pylint, AST parsing
• Rust - Clippy, AST parsing
• C# - dotnet format analyzers
• Go - golangci-lint (custom integration)
• Java, PHP, Ruby, C++ - Basic analysis
• And more...

---

🏗️ Development

Prerequisites

• Node.js 20+
• pnpm 9+

Setup

git clone https://github.com/zxcloli666/AI-Code-Review.git
cd AI-Code-Review
pnpm install
pnpm build

Project Structure

src/
├── index.ts              # Entry point
├── types/                # TypeScript types
├── ai/                   # AI client & prompts
│   ├── client.ts         # OpenAI/compatible client
│   ├── prompts.ts        # System & user prompts
│   └── tools-registry.ts # 14 AI tools registry
├── analysis/             # Code analysis engines
│   ├── ast-parser.ts     # AST parsing (Babel, TS, Acorn)
│   ├── linter-runner.ts  # Multi-language linter runner
│   └── call-graph.ts     # Dependency & call graph analysis
├── chunking/             # Smart chunking strategies
├── github/               # GitHub API integration
├── stats/                # Visualization & statistics
└── utils/                # Utilities

Testing

pnpm run build
pnpm run type-check

Create a test repository, set up the workflow, and verify the output.

---

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

1. Fork the repository
2. Create feature branch: git checkout -b feature/amazing-feature
3. Commit changes: git commit -m 'feat: add amazing feature'
4. Push to branch: git push origin feature/amazing-feature
5. Open Pull Request

---

📝 License

MIT License - see LICENSE for details.

---

🙏 Acknowledgments

• OpenAI - GPT-5, O3 models
• Anthropic - Claude Opus 4
• Google - Gemini 2.5 Pro
• GitHub Actions
• @babel/parser - JavaScript/TypeScript AST
• @typescript-eslint/parser - TypeScript analysis
• Rust Clippy - Rust linting
• .NET Roslyn Analyzers - C# analysis

---

⭐ Support This Project

If you find AI Code Review useful, please:

• ⭐ Star this repository
• 🐛 Report bugs or request features
• 🔀 Share with your team
• 💬 Join discussions

---

Made with ❤️ by zxcloli666

⬆ Back to top

---